To understand convolutional networks and visualize what hidden units in different layers are computing, for example a hidden unit in layer 1, Zeiler and Fergus. (2013) find
